Transcription API for clinics and medical practices
Consult notes, research interviews, telehealth recordings and dictated letters as text — per minute, with no per-clinician subscription. Read the HIPAA section before you upload patient audio; it says plainly what we do not offer.
- Step 1Dictated letter, consult recording, research interview
- Step 2POST /api/v1/transcribe · language + diarize
- Step 3Transcript · summary · DOCX for the record
- Step 4Letter drafted, note filed, interview coded
From recording to business result: one POST, one poll, every output format from the same job.
Three things a medical business actually does with it
Each priced at the published rate — $8 per 1,000 minutes — about $0.48 an audio hour — as credits that never expire.
A clinician dictates referral and discharge letters into their phone between patients. Each recording posts to the API; the typed letter is waiting in the practice's document system by the end of the session, ready to review and sign.
A qualitative study with 60 recorded interviews. Batch-submit the folder; every interview comes back speaker-labelled with timestamps, so coding in NVivo or Excel starts the same week rather than after a transcription contract.
With patient consent, a recorded telehealth consult is transcribed and summarised. The clinician edits a summary instead of writing a note from memory — the recording itself is deleted on the schedule you set.
What you can rely on — and what we are not
Whipscribe is not a HIPAA business associate and does not sign BAAs. Do not upload protected health information under US law unless your own compliance review says the safeguards below meet your obligation. Dictated letters with identifiers removed, de-identified research audio and non-US practices are the common fits.
Encrypted storage, transcription on GPUs we own, no third-party model, no training on your audio, retention of 30 days free or 365 paid, and deletion on demand from any file page.
99+ languages, auto-detected. Medical vocabulary is where any general model is weakest — check a sample of your own dictation before you commit, using the free first transcript.
The whole integration
Two calls. No SDK to learn, no model to host, no GPU to rent.
# submit — returns a job id immediately curl -X POST https://whipscribe.com/api/v1/transcribe/url \ -H "X-API-Key: $KEY" \ -H "Idempotency-Key: medical-$RECORDING_ID" \ -H "Content-Type: application/json" \ -d '{"url":"https://example.com/recording.mp3","diarize":true,"word_timestamps":true}' → 202 {"job_id":"a1b2…","status":"queued"} # poll, then collect — every format from the one job curl "https://whipscribe.com/api/v1/jobs/$JOB" -H "X-API-Key: $KEY" curl "https://whipscribe.com/api/v1/jobs/$JOB/result?format=docx" -H "X-API-Key: $KEY" curl "https://whipscribe.com/api/v1/jobs/$JOB/insights" -H "X-API-Key: $KEY" # summary · quotes · topics
Wire it into what you already use
What it replaces or sits beside
Vendors medical teams compare us with — each has a side-by-side page.
Try it on a real recording before you build anything
Drop a file from your own business. First transcript free at any length, no card — that is how to check accuracy on your audio before writing code.
Questions medical teams ask
Is Whipscribe HIPAA compliant?
No. Whipscribe is not a HIPAA-covered business associate and does not offer a BAA. We say this plainly because the alternative is a clinic finding out after the fact. If your use involves PHI, your compliance review decides; the encryption, own-GPU processing, no-training and delete-on-demand controls are what it has to work with.
Can it handle medical terminology?
It is a general Whisper-family model, not a medical-specialty one. It handles common clinical vocabulary well and drug names and specialist terms less consistently. Run a real dictation through the free first transcript before deciding.
Can each clinician use it without a seat licence?
Yes. There are no seats. One API key for the practice, billed per minute of audio, regardless of how many people dictate.
What does transcription cost for a medical business?
$8 per 1,000 minutes — about $0.48 an audio hour — as credits that never expire. First transcript free at any length. There are no seats and no monthly minimum: a 3-hour recording is about $1.44, two hundred half-hour calls about $48, a thousand-hour archive about $480 once.
Is there a free tier on the API?
No. The API has no free tier — a key needs a positive balance ($50 minimum, spendable on transcription). The free part is the web app: your first transcript there is free at any length, which is how to check accuracy on your own audio before paying anything.
How do we get an API key?
Self-serve: sign in, add credit, create the key at /apis/keys. No email, no sales call. Up to 10 active keys per account; rotate and revoke yourself.
How do results come back?
Submit, store the job id, poll GET /api/v1/jobs/{id} until done, then pull txt, json, srt, vtt or docx from the same job. Self-serve keys collect by polling; signed webhooks are an Enterprise feature.
Where is our audio processed?
On GPUs we own, in a private, secured cloud. It is never forwarded to OpenAI or any third-party model, and never used for training.
How long is audio kept?
30 days on the free plan, 365 on paid, and any file can be deleted on demand from its page. Wire deletion into your own retention rule if you need it automatic.
Does it label who is speaking?
Yes, when you enable diarize at submit time. Each speaker change is labelled; map the labels to names in your own code once.
What languages does it handle?
99+ languages, auto-detected, or set explicitly. Set it when you know it — auto-detect on a short or noisy clip is the most common cause of a wrong-language transcript.
Can we automate it with tools we already use?
Yes — Zapier, Make, n8n, Airtable, Drive, Dropbox, S3, Zoom, Slack, Notion. Each integration page carries the platform's real step timeout and the submit-then-poll shape that survives it.
What about very long or very large recordings?
Files go up to several GB and are submitted whole, so timestamps stay coherent — nothing is chunked. Hand us a URL and let us fetch it rather than streaming it through a no-code platform.
Do we get a summary as well as the transcript?
Yes. GET /api/v1/jobs/{id}/insights returns a summary, key quotes, topics and speakers for the same job, at no extra charge.
Can we try it before building anything?
Yes — the widget on this page. Drop a real recording from your business; the first transcript is free at any length, no card, and you will know whether the accuracy is there before writing a line of code.
Demand behind this page — GSC 90d: 155 medical queries, 632 impressions, position 42.7, 0 clicks — 'nuance dragon medical dictation software', 'dragon medical one enterprise licensing', 'dragon physician dictation', 'ava scribe'. People pricing per-seat dictation software, landing on our vendor pages, finding no business page.